Re: [RFC 0/8] Define coherent device memory node

[Anshuman Khandual <khandual@linux.vnet.ibm.com>]

On 10/27/2016 10:08 AM, Anshuman Khandual wrote:
> On 10/26/2016 09:32 PM, Jerome Glisse wrote:
>> On Wed, Oct 26, 2016 at 04:43:10PM +0530, Anshuman Khandual wrote:
>>> On 10/26/2016 12:22 AM, Jerome Glisse wrote:
>>>> On Tue, Oct 25, 2016 at 11:01:08PM +0530, Aneesh Kumar K.V wrote:
>>>>> Jerome Glisse <j.glisse@gmail.com> writes:
>>>>>
>>>>>> On Tue, Oct 25, 2016 at 10:29:38AM +0530, Aneesh Kumar K.V wrote:
>>>>>>> Jerome Glisse <j.glisse@gmail.com> writes:
>>>>>>>> On Mon, Oct 24, 2016 at 10:01:49AM +0530, Anshuman Khandual wrote:
>>>>>>
>>>>>> [...]
>>>>>>
>>>>>>>> You can take a look at hmm-v13 if you want to see how i do non LRU page
>>>>>>>> migration. While i put most of the migration code inside hmm_migrate.c it
>>>>>>>> could easily be move to migrate.c without hmm_ prefix.
>>>>>>>>
>>>>>>>> There is 2 missing piece with existing migrate code. First is to put memory
>>>>>>>> allocation for destination under control of who call the migrate code. Second
>>>>>>>> is to allow offloading the copy operation to device (ie not use the CPU to
>>>>>>>> copy data).
>>>>>>>>
>>>>>>>> I believe same requirement also make sense for platform you are targeting.
>>>>>>>> Thus same code can be use.
>>>>>>>>
>>>>>>>> hmm-v13 https://cgit.freedesktop.org/~glisse/linux/log/?h=hmm-v13
>>>>>>>>
>>>>>>>> I haven't posted this patchset yet because we are doing some modifications
>>>>>>>> to the device driver API to accomodate some new features. But the ZONE_DEVICE
>>>>>>>> changes and the overall migration code will stay the same more or less (i have
>>>>>>>> patches that move it to migrate.c and share more code with existing migrate
>>>>>>>> code).
>>>>>>>>
>>>>>>>> If you think i missed anything about lru and page cache please point it to
>>>>>>>> me. Because when i audited code for that i didn't see any road block with
>>>>>>>> the few fs i was looking at (ext4, xfs and core page cache code).
>>>>>>>>
>>>>>>>
>>>>>>> The other restriction around ZONE_DEVICE is, it is not a managed zone.
>>>>>>> That prevents any direct allocation from coherent device by application.
>>>>>>> ie, we would like to force allocation from coherent device using
>>>>>>> interface like mbind(MPOL_BIND..) . Is that possible with ZONE_DEVICE ?
>>>>>>
>>>>>> To achieve this we rely on device fault code path ie when device take a page fault
>>>>>> with help of HMM it will use existing memory if any for fault address but if CPU
>>>>>> page table is empty (and it is not file back vma because of readback) then device
>>>>>> can directly allocate device memory and HMM will update CPU page table to point to
>>>>>> newly allocated device memory.
>>>>>>
>>>>>
>>>>> That is ok if the device touch the page first. What if we want the
>>>>> allocation touched first by cpu to come from GPU ?. Should we always
>>>>> depend on GPU driver to migrate such pages later from system RAM to GPU
>>>>> memory ?
>>>>>
>>>>
>>>> I am not sure what kind of workload would rather have every first CPU access for
>>>> a range to use device memory. So no my code does not handle that and it is pointless
>>>> for it as CPU can not access device memory for me.
>>>>
>>>> That said nothing forbid to add support for ZONE_DEVICE with mbind() like syscall.
>>>> Thought my personnal preference would still be to avoid use of such generic syscall
>>>> but have device driver set allocation policy through its own userspace API (device
>>>> driver could reuse internal of mbind() to achieve the end result).
>>>>
>>>> I am not saying that eveything you want to do is doable now with HMM but, nothing
>>>> preclude achieving what you want to achieve using ZONE_DEVICE. I really don't think
>>>> any of the existing mm mechanism (kswapd, lru, numa, ...) are nice fit and can be reuse
>>>> with device memory.
>>>>
>>>> Each device is so different from the other that i don't believe in a one API fit all.
>>>> The drm GPU subsystem of the kernel is a testimony of how little can be share when it
>>>> comes to GPU. The only common code is modesetting. Everything that deals with how to
>>>> use GPU to compute stuff is per device and most of the logic is in userspace. So i do
>>>> not see any commonality that could be abstracted at syscall level. I would rather let
>>>> device driver stack (kernel and userspace) take such decision and have the higher level
>>>> API (OpenCL, Cuda, C++17, ...) expose something that make sense for each of them.
>>>> Programmer target those high level API and they intend to use the mechanism each offer
>>>> to manage memory and memory placement. I would say forcing them to use a second linux
>>>> specific API to achieve the latter is wrong, at lest for now.
>>>>
>>>> So in the end if the mbind() syscall is done by the userspace side of the device driver
>>>> then why not just having the device driver communicate this through its own kernel
>>>> API (which can be much more expressive than what standardize syscall offers). I would
>>>> rather avoid making change to any syscall for now.
>>>>
>>>> If latter, down the road, once the userspace ecosystem stabilize, we see that there
>>>> is a good level at which we can abstract memory policy for enough devices then and
>>>> only then it would make sense to either introduce new syscall or grow/modify existing
>>>> one. Right now i fear we could only make bad decision that we would regret down the
>>>> road.
>>>>
>>>> I think we can achieve memory device support with the minimum amount of changes to mm
>>>> code and existing mm mechanism. Using ZONE_DEVICE already make sure that such memory
>>>> is kept out of most mm mechanism and hence avoid all the changes you had to make for
>>>> CDM node. It just looks a better fit from my point of view. I think it is worth
>>>> considering for your use case too. I am sure folks writting the device driver would
>>>> rather share more code between platform with grown up bus system (CAPI, CCIX, ...)
>>>> vs platform with kid bus system (PCIE let's forget about PCI and ISA :))
>>>
>>> Because of coherent access between the CPU and the device, the intention is to use
>>> the same buffer (VMA) accessed between CPU and device interchangeably through out
>>> the run time of the application depending upon which side is accessing more and
>>> how much of performance benefit it will provide after the migration. Now driver
>>> managed memory is non LRU (whether we use ZONE_DEVICE or not) and we had issues
>>> migrating non LRU pages mapped in user space. I am not sure whether Minchan had
>>> changed the basic non LRU migration enablement code to support mapped non LRU
>>> pages well. So in that case how we are going to migrate back and forth between
>>> system RAM and device memory ?
>>
>> In my patchset there is no policy, it is all under device driver control which
>> decide what range of memory is migrated and when. I think only device driver as
>> proper knowledge to make such decision. By coalescing data from GPU counters and
>> request from application made through the uppler level programming API like
>> Cuda.
>>
> 
> Right, I understand that. But what I pointed out here is that there are problems
> now migrating user mapped pages back and forth between LRU system RAM memory and
> non LRU device memory which is yet to be solved. Because you are proposing a non
> LRU based design with ZONE_DEVICE, how we are solving/working around these
> problems for bi-directional migration ?

Let me elaborate on this bit more. Before non LRU migration support patch series
from Minchan, it was not possible to migrate non LRU pages which are generally
driver managed through migrate_pages interface. This was affecting the ability
to do compaction on platforms which has a large share of non LRU pages. That series
actually solved the migration problem and allowed compaction. But it still did not
solve the migration problem for non LRU *user mapped* pages. So if the non LRU pages
are mapped into a process's page table and being accessed from user space, it can
not be moved using migrate_pages interface.

Minchan had a draft solution for that problem which is still hosted here. On his
suggestion I had tried this solution but still faced some other problems during
mapped pages migration. (NOTE: IIRC this was not posted in the community)

git://git.kernel.org/pub/scm/linux/kernel/git/minchan/linux.git with the following
branch (non-lru-mapped-v1r2-v4.7-rc4-mmotm-2016-06-24-15-53) 

As I had mentioned earlier, we intend to support all possible migrations between
system RAM (LRU) and device memory (Non LRU) for user space mapped pages.

(1) System RAM (Anon mapping) --> Device memory, back and forth many times
(2) System RAM (File mapping) --> Device memory, back and forth many times

This is not happening now with non LRU pages. Here are some of reasons but before
that some notes.

* Driver initiates all the migrations
* Driver does the isolation of pages
* Driver puts the isolated pages in a linked list
* Driver passes the linked list to migrate_pages interface for migration
* IIRC isolation of non LRU pages happens through page->as->aops->isolate_page call
* If migration fails, call page->as->aops->putback_page to give the page back to the
  device driver

1. queue_pages_range() currently does not work with non LRU pages, needs to be fixed

2. After a successful migration from non LRU device memory to LRU system RAM, the non
   LRU will be freed back. Right now migrate_pages releases these pages to buddy, but
   in this situation we need the pages to be given back to the driver instead. Hence
   migrate_pages needs to be changed to accommodate this.

3. After LRU system RAM to non LRU device migration for a mapped page, does the new
   page (which came from device memory) will be part of core MM LRU either for Anon
   or File mapping ?

4. After LRU (Anon mapped) system RAM to non LRU device migration for a mapped page,
   how we are going to store "address_space->address_space_operations" and "Anon VMA
   Chain" reverse mapping information both on the page->mapping element ?

5. After LRU (File mapped) system RAM to non LRU device migration for a mapped page,
   how we are going to store "address_space->address_space_operations" of the device
   driver and radix tree based reverse mapping information for the existing file
   mapping both on the same page->mapping element ?

6. IIRC, it was not possible to retain the non LRU identify (page->as->aops which will
   defined inside the device driver) and the reverse mapping information (either anon
   or file mapping) together after first round of migration. This non LRU identity needs
   to be retained continuously if we ever need to return this page to device driver after
   successful migration to system RAM or for isolation/putback purpose or something else.

All the reasons explained above was preventing a continuous ping-pong scheme of migration
between system RAM LRU buddy pages and device memory non LRU pages which is one of the
primary requirements for exploiting coherent device memory. Do you think we can solve these
problems with ZONE_DEVICE and HMM framework ?